Skip to content

Install missing BCM43602 board NVRAM on MacBookPro13,3 - #11017

Open
justin-schroeder wants to merge 3 commits into
omacom:quattrofrom
justin-schroeder:fix/apple-bcm43602-board-nvram
Open

justin-schroeder wants to merge 3 commits into
omacom:quattrofrom
justin-schroeder:fix/apple-bcm43602-board-nvram

Conversation

@justin-schroeder

Copy link
Copy Markdown

Install missing BCM43602 board NVRAM on MacBookPro13,3

On a MacBookPro13,3 with BCM43602 [14e4:43ba], the stock binary firmware initialized without board NVRAM but only discovered 2.4 GHz networks and reported failed WPA2 handshakes as incorrect passwords. Installing the board configuration and reloading brcmfmac restored 5 GHz discovery and a successful connection at 5240 MHz, with signal −39 dBm.

Add a hardware setup leaf after the existing brcmfmac supplicant workaround. Match Apple DMI, MacBookPro13,3, and PCI/subsystem IDs 14e4:43ba / 106b:015a; the same chip in other models is not sufficient evidence of compatible board calibration. Fetch the tested community NVRAM from an immutable URL and verify SHA-256 before writing a model-specific firmware file. Preserve existing generic/model-specific overrides, including compressed files and symlinks. Substitute the interface MAC; if absent or the shared Broadcom placeholder, generate a persistent locally administered unicast address. Do not reload the adapter during installation.

Source: Mike Ratcliffe’s BCM43602 board configuration, pinned file blob bd7af7c6f01df3ccee1471d36cb8df15d618a6aa, SHA-256 b109f3e6663b0e888c2559e36f7e0109f2a3a6b9765786d11f849f16d4b32d06. The data is downloaded rather than vendored; matching installations require network access to the pinned source. Download/checksum failures stop the leaf without installing an unverified file.

Validation:

  • Real hardware: MacBookPro13,3, BCM43602 rev 02; Linux 7.1.9-arch1-2; firmware 7.35.177.61; NetworkManager 1.58.1-1. Verified multiple 5 GHz scan results and successful 5 GHz connection after adding the same NVRAM data and reloading brcmfmac_wcc/brcmfmac. The pre-existing feature_disable=0x82000 setting was retained and applied on reload.
  • Focused sandbox tests cover hardware exclusions, existing overrides, MAC handling, repeat runs, failed downloads and checksum mismatches.
  • Bash syntax and git whitespace checks pass.
  • Full ./test/all: CLI suite passed; 5 of 237 shell test files failed. config-test.sh, snapper-test.sh, and unowned-system-paths-test.sh require the absent companion omarchy-pkgs checkout; launch-about-test.sh fails its roomy-window animation assertion. These four failures reproduce on untouched upstream 5ead8705. locate-test.sh tried to decode ignored Python bytecode generated in bin/pycache by earlier tests; it passes after removing that generated cache. The new Wi-Fi suite passes. No unrelated test changes are included.

Clean ISO installation, reboot persistence, successful 2.4 GHz authentication after the change, other Mac models and long-term stability have not been validated. This PR adds fresh-install hardware setup; it does not add a migration or change WPA settings, regulatory settings, binary firmware, or suspend behavior.

@omarchybot omarchybot added bug Something isn't working compatibility Make Omarchy work better on everything! labels Sep 27, 2026
omarchybot and others added 2 commits October 2, 2026 13:03
run_logged returns the leaf's exit status and omarchy-apply-hardware runs under set -e, so a failed download or checksum stopped every hardware leaf after this one and failed omarchy-apply-system. An offline install on a MacBookPro13,3, the machine whose Wi-Fi this is fixing, never finished hardware setup. Skip the board file instead and leave the stock firmware in place.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
brcmfmac requests the model-specific board file before the generic one, so a generic brcmfmac43602-pcie.txt an administrator placed in /usr/lib/firmware/updates/brcm would be overridden by the file this leaf installs. Look for both names there as well as in brcm/.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex Medium <noreply@openai.com>
@omarchybot

Copy link
Copy Markdown
Collaborator

Thanks for the careful write-up, @justin-schroeder. I reviewed this at a4980020 and pushed two fixes to your branch, now at 7582832940781f316284a55dc16ec99387e600ba. The review was by Claude Opus 5.5, with Codex Medium as a second opinion over two rounds; it found the second defect below, and its last round on the new head found nothing further.

Pushed:

  • 614801fd — a failed download stopped the whole install. run_logged passes the leaf's exit status back and omarchy-apply-hardware runs under set -e, so on a MacBookPro13,3 with no network during installation the curl failure skipped every hardware leaf after this one and failed omarchy-apply-system. I reproduced that on a worker: the next leaf never ran and setup exited 6. A failed download or checksum now prints a message, leaves the stock firmware in place, and returns 0. The test asserts that and fails against the original leaf.
  • 75828329 — an override under firmware/updates was ignored. brcmfmac requests the model-specific file before the generic one, so a generic brcmfmac43602-pcie.txt an administrator had put in /usr/lib/firmware/updates/brcm/ would have been overridden by the file this installs. The leaf now looks in updates/brcm as well as brcm, and the test covers both.

Checked: on a disposable Omarchy worker, ./test/cli and test/shell.d/apple-bcm43602-nvram-test.sh pass on the new head, and the pinned gist still matches its SHA-256. A worker has no BCM43602, so the Wi-Fi change itself was read, not run. The DMI and PCI gate is as narrow as it says, and brcmfmac43602-pcie.Apple Inc.-MacBookPro13,3.txt is the name brcmfmac builds from DMI.

What it waits on: a test of this head on a real MacBookPro13,3, on both bands. Your own result covers 5 GHz. 2.4 GHz is the open question, because the #7487 thread has a 13,3 (106b:015a) where a different community calibration fixed 5 GHz and stopped 2.4 GHz from associating. This file is a close relative of the 14,2/14,3 calibration in #7671 (boardtype=0x61b; it differs only in boardflags3 and swctrlmap_2g). That confirmation has to come from someone other than the pull request's author. @andyholst, you reported #11613 on this model; if you can, please try it:

git clone -b fix/apple-bcm43602-board-nvram https://github.com/justin-schroeder/omarchy.git /tmp/omarchy-11017
git -C /tmp/omarchy-11017 checkout 7582832940781f316284a55dc16ec99387e600ba
sudo bash -e -c 'source /tmp/omarchy-11017/install/hardware/apple/fix-bcm43602-nvram.sh'
sudo reboot

The leaf stops without changing anything if any brcmfmac43602-pcie*.txt already exists in /usr/lib/firmware/brcm or /usr/lib/firmware/updates/brcm, so move one aside first if you installed one. After rebooting, please report iw dev and iw dev <interface> link while connected to a 5 GHz network, and again while connected to a 2.4 GHz WPA2 network.

For the maintainer, not the author:

  • Draft Stabilize MacBookPro13,3 hardware support #10141 also installs a MacBookPro13,3 calibration (bugzilla attachment 285753, boardtype=0x073e), as part of a larger 13,3 bundle, so the two compete for this machine's Wi-Fi. This one is the smaller of the two and the one I would carry forward; the second opinion, asked separately, picked the same, though its independence is not guaranteed. Neither is proven on 2.4 GHz.
  • This writes an unpackaged file into /usr/lib/firmware/brcm, which test/shell.d/unowned-system-paths-test.sh warns against: if a package ever ships that name, pacman refuses the upgrade. The scan doesn't flag it because the path is built from a variable. Install BCM43602 NVRAM so 2017 Touch Bar Macs see 5 GHz Wi-Fi #7671 uses /usr/lib/firmware/updates/brcm and records it in the test instead.
  • It runs at install time only, with no migration, so existing 13,3 installs don't get it. It also downloads from a personal gist during installation. The hash pins what is downloaded, but an offline install now just skips it.

@greptile-apps

greptile-apps Bot commented Oct 2, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerConfidence Score: 4/5

[High risk] Adds hardware setup script that downloads and installs firmware.

The PR is not ready to merge while supported offline installations can complete without applying the intended Wi-Fi fix.

Findings

  1. P1 Offline installs skip the fix ▶
  2. P2 Pinned digest goes untested ▶

Summary

Adds a MacBookPro13,3-specific hardware leaf that verifies and installs BCM43602 board NVRAM, preserves existing overrides, and sets a persistent MAC address. Adds focused tests for matching, installation, and failure cases.

Reviews (1) · Last reviewed commit: "Respect BCM43602 NVRAM overrides under f..."

Comment on lines +39 to +42
! echo "b109f3e6663b0e888c2559e36f7e0109f2a3a6b9765786d11f849f16d4b32d06 $scratch/nvram" | sha256sum --check --status; then
echo "Could not fetch verified MacBookPro13,3 BCM43602 board NVRAM; skipping"
return 0
fi

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Offline installs skip the fix If a matching Mac is installed without access to the pinned download, this branch reports success without installing the board NVRAM. Hardware setup is not automatically retried, so the machine boots without the Wi-Fi fix. The installer supports offline package installation; this leaf needs a way to complete the fix in that path or retry it later.

Knowledge Base Used: Install stages and package sets

Comment on lines +58 to +59
def invoke():
return subprocess.run(["bash", "-e", "-c", 'source "$1"', "bash", str(script)],

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Pinned digest goes untested The test replaces the production SHA-256 with the hash of its own fixture. It would still pass if the pinned digest were mistyped or the remote payload changed, even though installation would then fail verification and skip the board file. Check the actual pinned artifact against its digest as well.

@monomyth

monomyth commented Oct 8, 2026

Copy link
Copy Markdown

A clarification on the calibration comparison in the review: there is historical 2.4 GHz association evidence for the 285753-based configuration carried by #10141.

Our September 20 report records a MacBookPro13,3 / BCM43602 subsystem 106b:015a connected at 2412 MHz on September 16, running custom Linux 7.2.5-4-omarchy-mbp133. The installed calibration files were unchanged, SHA-256 a1a13fe89cf570ce94237cd27de36030584e8ec7084c3447c20c4fd668411dba. The same report links the separate September 15 5600 MHz / channel 120 association on custom 7.2.4-arch1-2-mbp133.

These were different boots and kernel versions, not a controlled same-kernel dual-band trial or validation of #10141's current installer head. We have not tested this PR's different calibration or head 7582832940781f316284a55dc16ec99387e600ba. The observations do not establish which calibration is preferable, prove long-term reliability, or satisfy this PR's requested dual-band hardware test.

So "neither is proven on 2.4 GHz" should distinguish missing current-head validation from the existing association evidence above. This is a link to prior evidence only; no new radio reload, firmware change, reboot or hardware test was performed for this reply.

@omarchybot

Copy link
Copy Markdown
Collaborator

I took another look at 75828329, the same head as last time, because of the Greptile review and @monomyth's note. Claude Opus 5.5 did the review, and Codex Medium gave a second opinion. Neither found anything new wrong with the code. The PR still merges cleanly onto current quattro. One part of my earlier comment needs correcting.

The competing Wi-Fi fix. @monomyth is right that I overstated it when I said neither calibration was proven on 2.4 GHz. The 285753-based file in #10141 has been seen connected at 2412 MHz on one boot and at 5600 MHz on another. Those were custom kernels, not a controlled test, but they are positive results on both bands. I was also wrong to call this PR the one to carry forward. When two fixes for the same bug compete, this project lets the older one win unless it has a specific defect. #10141's Wi-Fi part is older (opened 4 September; this one 9 September), and nobody has shown a defect in it. Its draft status and the unrelated Radeon, Touch Bar and suspend work bundled with it hold back #10141 as a whole. They are not faults in its Wi-Fi fix. Codex, asked separately, came to the same conclusion, though its independence is not guaranteed. Which 13,3 calibration lands is for the maintainer to decide.

Greptile's two findings:

  • Offline installs skip the fix: that is accurate, and it was deliberate. Failing the leaf would stop every hardware step after it, and nothing retries it later. The alternative is to ship the file in the repository, as Install BCM43602 NVRAM so 2017 Touch Bar Macs see 5 GHz Wi-Fi #7671 and Stabilize MacBookPro13,3 hardware support #10141 do, and that decision belongs to the maintainer.
  • Pinned digest goes untested: this is only partly right. The test replaces that exact digest string, so a typo in the script's digest makes the install test fail. Codex pointed this out. What the suite cannot check is whether the digest matches the real gist file. It did match when I checked on 2 October.

What it waits on: the maintainer, to choose between this PR and the Wi-Fi part of #10141 for the MacBookPro13,3. Whichever one is chosen still needs someone other than its author to test it on real 13,3 hardware on both bands.

@jandrusk

Copy link
Copy Markdown

Addresses the missing board NVRAM on MacBookPro13,3 (14e4:43ba). Please coordinate with #7671 (NVRAM install) and #10332 (Wi-Fi after suspend) so the Mac BCM43602 story lands as one coherent path rather than three overlapping installs.

@omarchybot

Copy link
Copy Markdown
Collaborator

On @jandrusk's point about #7671 and #10332: I checked how the three fit together at this head (75828329, unchanged since my last comment). They are not three installs of the same thing. Only one line conflicts between them. Claude Opus 5.5 did the review, and Codex Medium gave a second opinion that agreed. Its independence is not guaranteed.

  • Install BCM43602 NVRAM so 2017 Touch Bar Macs see 5 GHz Wi-Fi #7671 never runs on this machine. Its gate is MacBookPro14,2 and 14,3 only, and it excludes 13,3 on purpose. This PR is gated to MacBookPro13,3 with subsystem 106b:015a, so no Mac gets both calibrations. If a generic brcmfmac43602-pcie.txt is already present in either firmware directory, this leaf leaves it alone.
  • Fix Wi-Fi dying after suspend on BCM43602 MacBooks #10332 solves a different problem. It adds a sleep hook that reloads brcmfmac around suspend, and it writes no calibration. On a 13,3 the two work together: the reload picks up whichever NVRAM is installed.
  • All three add a run_logged line at the same spot in install/hardware/all.sh, right after fix-brcmfmac-supplicant.sh. Whichever lands second gets a one-line conflict, and the fix is to keep all three lines. Each one still merges cleanly onto current quattro.

The overlap that does need a decision is the one in my earlier comment. Draft #10141 installs a different MacBookPro13,3 calibration (boardtype=0x073e). Whichever lands first suppresses the other, because both leaves skip when a board file is already present. That choice is still the maintainer's. Whichever one is chosen still needs someone other than its author to test it on both bands on a real 13,3.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working compatibility Make Omarchy work better on everything!

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants